Lightnews — Scholar-powered news

Liran Tal @lirantal.com · 12h

it seems timely to share my long-time comprehensive list of Node.js CLI applications best practices on GitHub: github.com/lirantal/nod...

Liran Tal @lirantal.com · 15h

security is so simple

2

Liran Tal @lirantal.com · 21h

I like how this new feature to detect pinned versions of model context protocol servers has turned out

now if you use `npx ls-mcp` then it gives you insights whether an MCP server can do a rug pull on you so you'd know if you have a security exposure to that kind of incident

Liran Tal @lirantal.com · 1d

new summary section for the ls-mcp npx package that detects running Model Context Protocol servers across your AI agent apps, kinda slick no?

Liran Tal @lirantal.com · 1d

oh you're using MCP but your MCP servers are using implicit "latest" versions... aha.. rug pull incoming?

the new ls-mcp feature detects if this is the case, just use it and check you're not going to crash into a security incident:

$ npx ls-mcp

1

Reposted by Liran Tal

Kamran Ayub @kamranicus.com · 1d

Community Highlights 💖

- @lirantal.com shares supply chain security solutions with DevContainers and 1Password
- Vincent Rolfs explains why the TypeScript codebase uses var statements
- Paul points out where type safety can be an illusion
- @t3.gg reacts to the Moving Off TypeScript blog post

1 1

Liran Tal @lirantal.com · 1d

don't expose yourself to supply chain attacks and security incidents on npm

get npq today

2

Liran Tal @lirantal.com · 1d

great walkthrough of using secrets vaults and password managers to avoid the .env / dotenv credentials leak in supply chain attacks and other ways they get leaked:

The case against secrets in .env files

Most developers rely on .env files to store secrets like API keys, database passwords, and tokens. But what if I told you this common practice can leave you wide open to attacks? In this video, I break down why storing secrets in a .env file is dangerous, how attackers can exploit it, and what safer alternatives you should be using instead. Use Snyk for free to find and fix security issues in your applications today! https://snyk.co/ugLYn ✍️ Resources ✍️ - Jira MCP Hijack Post: https://x.com/mbrg0/status/1953932780855013682 - Crypto Extension Hack Post: https://x.com/0xzak/status/1955265807807545763 - Weaponizing AI Coding Agents Post: https://snyk.co/ujcke - Shai-Hulud Supply Chain Attack Post: https://snyk.co/ujckf - Compromised Open Source Maintainer Post: https://snyk.co/ujckg ⏲️ Chapters ⏲️ 00:00 - Intro 01:03 - Why .env files are bad 01:56 - Safer alternatives 02:39 - Doppler demo 04:56 - How this works 07:16 - 1Password demo 10:00 - Why this is good 10:28 -...

www.youtube.com

1 4

Liran Tal @lirantal.com · 2d

I really like this talk from Human Layer own Dexter Horthy, especially the insights about flawed rules files (a la AGENT.md) and their compounding impact for flawed code

Advanced Context Engineering for Agents

Dexter Horthy, founder of Human Layer, shares what his team has learned about scaling coding agents in real-world software projects. He walks through why naive back-and-forth prompting fails, how spec-first development keeps teams aligned, and why “everything is context engineering.” From compaction strategies to subagents and planning workflows, he shows how intentional context management turns AI coding from prototypes into production. Chapters: 00:09 - The Origin of Context Engineering 00:46 - Key Talks and Insights from AI Engineering 01:45 - Challenges with AI in Complex Systems 03:12 - The Shift to Spec-First Development 04:03 - Advanced Context Engineering for Coding Agents 04:48 - Intentional Compassion in Context Management 05:45 - Optimizing Context Utilization 07:27 - The Role of Subagents in Context Control 08:48 - Frequent Intentional Compaction 11:00 - Practical Implementation and Workflow 11:12 - Case Study: Fixing a Rust Code Base 11:59 - Insights on Effective Coding Practices 12:44...

www.youtube.com

1 1

Liran Tal @lirantal.com · 2d

interesting benchmark regarding code security with LLM vs SAST security tools

I'm definitely curious to re-run this myself and figure out what these "false positives" are about with Snyk but nonetheless good to see Snyk performs better than the competition

4

Liran Tal @lirantal.com · 2d

if you have this MCP Server definition in your configuration file it shouldn't be a security risk because the token isn't saved in plain text

ls-mcp doesn't have a false positive detection, which is great

but in other cases, ls-mcp will warn you of credentials leak

Liran Tal @lirantal.com · 2d

thankful to GitHub for attributing to me for finding and disclosing the path traversal vulnerability in servst npm package

servst was last published 7 years ago and has virtually 0 downloads but hey, keeping you secure is my goal and I'm not stopping!

2

Liran Tal @lirantal.com · 3d

finally got GitHub to apply proper credits to my private-ip SSRF security disclosure CVE

3

Liran Tal @lirantal.com · 3d

I received good feedback on this beginner's guide to MCP architecture so re-posting it: snyk.io/articles/a-b...

A Beginner's Guide to Visually Understanding MCP Architecture | Snyk

A beginner's guide to understanding Model Context Protocol (MCP) architecture visually. Explore MCP Hosts, Clients, Servers, transport types, and how MCP extends LLMs beyond function calling & REST APIs.

snyk.io

1 3

Liran Tal @lirantal.com · 3d

you realize the new wave of AI agents security issues and MCP servers vulnerabilities are just....... plain old SQL injection, yes?

here's why
here's more about the why: labs.snyk.io/resources/se...

3 4

Liran Tal @lirantal.com · 3d

appreciate the shout-out Mahdy! thank you ❤️

I love educating developers about AI security and how MCP fits into mix. hope to see you again in the next Cursor user group night!

Liran Tal @lirantal.com · 4d

So did you already nerd out and played Snyk VulnVortex or not?! web-based platformer:

Vuln Vortex

Can you escape all the vulnerable npm packages? Try your luck with Vuln Vortex

www.vulnvortex.com

1

Liran Tal @lirantal.com · 4d

Well, a tip from me: you can set user role to whatever you want, it has no evident effect on the model from a security control point of view.

Liran Tal @lirantal.com · 4d

hello payload

1

Liran Tal @lirantal.com · 4d

Thank you Kamran ❤️

Liran Tal @lirantal.com · 4d

Will you be satisfied if I change "system" user role with "user" ?

1

Liran Tal @lirantal.com · 4d

I like those points, Mike, thanks! basically what you should be paying attention to is that message is user input variable that goes into an LLM, then that LLM response is directly used to send the HTTP response to the browser so -> XSS in this simple example

1 1

Liran Tal @lirantal.com · 4d

do you see the LLM centric security vulnerability here or should I point it out?

more context: labs.snyk.io/resources/se...

2 3

Liran Tal @lirantal.com · 4d

Given recent supply chain security incidents (namely S1ngularity with qix and Shai-Hulud) I wrote down a proposed workflow that would help mitigate the blast radius in case you were to be compromised: a DevContainers + 1Password workflow: www.nodejs-security.com/blog/mitigat...

This workflow isn't

Mitigate Supply Chain Security with DevContainers and 1Password for Node.js Local Development

How-to setup an isolated Node.js local development environment with VS Code DevContainers and 1Password to keep secrets out of your filesystem and avoid supply chain security incidents like shai-hulud, qix maintainer compromise and others.

www.nodejs-security.com

1 2

Liran Tal @lirantal.com · 4d

Thank you! this was a nice new find for me too across a bunch of MCP servers :-)